Re: [luv-talk] Statistical spam filters

On Fri, 3 May 2013, Jason White <jason@jasonjgw.net> wrote:
If you think that means you won't have users feeding unread messages from their "spam" folder into the statistical training system then you're wrong.
Proof by counter-example - great, thanks, and disappointing news it is.
Well it only really means that you can't implement shared statistical systems, or at least the shared part needs to have a low weight (like a maximum of 1 point out of 5+ needed for a spam classification). If a user wants to stuff things up for themself then it's their issue really. Of course the type of person who'll feed their anti-spam system in that way will probably do things that result in mail to the postmaster and/or abuse addresses more often than most users in your domain... -- My Main Blog http://etbe.coker.com.au/ My Documents Blog http://doc.coker.com.au/
participants (1)
-
Russell Coker